An integrated approach for multimedia information retrieval.

Retrieving information effectively from multimedia collections is an open research problem. This investigation addresses this retrieval problem in a generic way. A framework for multimedia information retrieval has been developed in which multiple feature representation and matching schemes can be incorporated. At the theoretical level this framework integrates multiple retrieval techniques, and the Dempster-Shafer theory of evidence combination is employed for this purpose. The integration of the exact and inexact match retrieval is thereby achieved in one computational model. The framework is supported at the architectural level by developing a flexible and extensible information retrieval (IR) development system, called FLAIR (Flexible Architecture for Information Retrieval). The purpose of the FLAIR is three-fold: to allow multiple feature representation schemes; to incorporate new retrieval techniques as and when they are available; and to integrate the exact and inexact matching schemes. The FLAIR software architecture provides abstractions of IR processes so that developers can use them in a mix-and-match fashion. These abstractions are extensible and hence can be used to develop and incorporate new functionalities into the system. The feasibility of this framework is demonstrated by building a photographic retrieval system called Epic. This system is based on the use of “spatial features”, which describe objects in the photograph and their locations as photographic features. We have shown that these features are useful in providing a natural querying tool for photograph retrieval. To evaluate the usefulness of this picture characterisation scheme we have employed a task-oriented, user-centred approach in which we used real graphic designers and observed how they interacted with the system. We have employed the concept of a simulated work task situation for creating a real-world information seeking scenario. The evaluation has shown the strengths and weaknesses of the spatial feature based retrieval system. This evaluation methodology has shown to be useful for evaluating interactive information retrieval systems. In summary, the contributions of this research are the following: a generic framework for multimedia information retrieval, a retrieval model in which multiple retrieval mechanisms can be integrated including database and information retrieval integration, a picture characterisation scheme, an extensible architecture for developing information retrieval applications, a prototype photograph retrieval system called Epic, and an evaluation methodology for interactive IR systems.

Grantor dc:publisher.institution

Robert Gordon University

Year dc:date.issued

1998

Author dc:creator

• Jose, Joemon M.

Advisor dc:contributor.advisor

• D. Harper

• Multimedia information retrieval
• Dempster-Shafer theory
• Flexible architecture for information retrieval (FLAIR)
• Photographic retrieval system (EPIC)
• Spatial feature retrieval system

Language dc:language

en

Identifier

oai:rgu-repository.worktribe.com:2807421
https://doi.org/10.48526/rgu-wt-2807421

OAI identifier oai:identifier

oai:rgu-repository.worktribe.com:2807421